The mathematician Bertrand Russell said, "There must be an ideal world, a sort of mathematician's paradise where everything happens as it does in textbooks." People involved with data-acquisition systems hope for a similar nirvana. Although data-acquisition (DAQ) systems often work flawlessly on their own, as soon as engineers connect them to sensors, they may experience problems: Noise appears unexpectedly and signals vary from those expected. Often, these degraded measurements occur due to subtle system issues that designers overlook.
Ground down: Multiple ground points (a) often cause ground loops that present instruments and DAQ systems with common-mode voltages. These voltages can produce unpredictable errors in measurements. Using a single ground point (b) in a star configuration helps eliminate such loops and currents.
Several years ago, while working with a DAQ system, I noticed the acquired data looked unusual. I had expected a small amount of noise, but the odd data I obtained raised concerns about the accuracy of the DAQ system. Taking apart the system and testing its individual subassemblies revealed no obvious problems. But after reassembling the system, the wildly varying results remained. Where did that system noise come from? A bit of investigation showed my design had succumbed to a dreaded ground loop.
The ground loop is one of several traps that fall into four broad categories—resistive, capacitive, inductive, and magnetic coupling—all of which can cause measurement errors. Good overall system design, though, can mitigate some of these traps.
Resistance is Futile
Resistive coupling comes about when designers think a ground is a ground is a ground. They forget that a ground connection can carry current and exhibit resistance. As a result, voltages can appear between ground points. Thus, improper grounding at separate points can form a ground loop that introduces unexpected voltages and currents in measurements.
Shield me: Electrical systems can't avoid capacitively couple noise. But a good shield can bleed off any accumulated charge so it doesn't affect signal-carrying conductors.
Designers may assume a system furnishes a solid, low-impedance ground, but that's not always the case. In practice, people use a variety of conductors, from a lab outlet to a building's steel structure or even a cold-water pipe, to provide a ground. To overcome grounding problems, designers can adopt a "star" topology that routes all ground signals to one common point through short low-impedance connections.
In my DAQ system, two ground connections ran to separate, remote circuit breakers. The long ground loop from the computer through the breaker box back to the DAQ system provided a path through which appreciable current could flow. A ground connection between the DAQ equipment and a computer completed the loop. Connecting all the equipment to a nearby common ground eliminated the loop and let me collect useful data.
Keep in mind that a ground shouldn't carry current; it simply provides a zero potential for all systems connected to it. Don't confuse a ground with a return path, part of a circuit that returns power to a source. (Unfortunately, designers can confuse the two and treat them equally.) In many cases, simply providing a single ground point will eliminate problems. But at times, the distance between equipment requires separate ground connections. In these cases, designers may have to use signal-conditioning techniques to break a ground loop and isolate a sensor from the DAQ equipment. A number of manufacturers offer modules that can electrically isolate sensors from measuring equipment, often using magnetic coupling rather than a galvanic connection.
After talking with the ADC manufacturer's application engineer, he suggested I place a 1-MÙ resistor between each differential input and ground. The resistors provided a ground path for bias currents and thus stabilized the ADC's readings. Unfortunately, the resistors reduced the impedance of the ADC's inputs, thus loading the sensor circuits slightly, but without affecting the final measurements. (Unconnected DAQ-system inputs can "float" at unknown voltages and affect measurements too, so always connect these inputs to a nearby analog ground.)
A Capacity for Noise
Capacitive coupling takes place through "stray," or mutual, capacitance inherent in the structure of electronic components and assemblies. Although designers can't eliminate stray capacitance, they can reduce its effects, usually by surrounding a signal-carrying conductor with a conductive shield. When properly grounded, the shield dissipates any charge built up by capacitive coupling. In effect, the outer conductor provides a Faraday shield around the internal wire or wires.
A proper shield connects to ground at only one point. This single, short connection provides a low-impedance path for any accumulated charge. Two grounds aren't better than one, though. Connecting the shield to ground at both ends of a cable establishes a ground loop. And any current that flows in a shield will induce a voltage in the center conductor or conductors. If you plan to run cables to several sensors, use a separate shielded cable for each sensor and connect each shield to the respective sensor ground. Do not simply connect all the shields to one ground.
When a sensor provides a differential signal, system designers should employ cable that supplies two conductors inside a shield. Never use the shield to provide a signal path or to carry power. Again, ensure only one ground exists between your equipment and the sensor.
Don't Talk to Others
Inductive coupling also can add noise to a signal. Often when a DAQ system connects to many sensors, signals travel through bundles of wires. But placing so many signals close together can cause an inductive effect called crosstalk, the addition of small amounts of nearby signals to each other.
If engineers cannot route signals through shielded cables, they often use twisted-pair wires. Inductive coupling takes place through loops, and minimizing the loop area reduces the coupling of noise. The twists—about 10 to 12 per foot—also subject both conductors, on average, to equal coupling, and a DAQ system can usually reject much of any common-mode noise that exists equally on two inputs.
Sometimes, engineers cannot avoid placing digital, power, and analog signals in the same bundles. In such cases, they strive to use a type of cable that segregates signal types.
You've probably seen the multiconductor flat cables that connect computers to other equipment. DAQ systems also may use this type of cable to connect with sensors or terminal blocks. Although simple flat cables handle digital signals well, engineers avoid it for low-voltage sensor applications. Instead, they choose a flat cable that provides twisted pairs of wires and an overall shield. These characteristics can help reduce the effects of capacitive and inductive coupling. Instrumentation-grade flat cable comes from many vendors in various wire gauges and configurations.
Block that Field
Nearby magnetic fields also may disrupt sensitive measurements. Low-frequency(&1 kHz) fields can result from welders, motors, transformers, solenoids, and other devices. Unfortunately, the thin electrical shielding used to reduce capacitive coupling has little effect on noise caused by magnetic fields. Reducing the influence of magnetic signals at 50 or 60 Hz can require thick shields. To reduce a 60 Hz magnetic signal by 63% (1-e-1) requires 0.34 inch of copper, 0.034 inch of steel, or 0.43 inch of aluminum. Remember though, any material loses its effectiveness as a shield when magnetically saturated.
Bensenville, IL-based Magnetic Shield Corp. (www.magnetic-shielding.com) offers an inexpensive kit ($130) that includes samples of shielding materials, a shielding slide-rule calculator, and a magnetic-field probe. If you face a difficult magnetic-field interference problem, the kit may help you decide how to overcome it.